{
  "fetched_at": "2026-10-19T09:14:13Z",
  "source_url": "https://www.kaggle.com/competitions/coms4771-fall2021-regression-competition/data",
  "pages": [
    {
      "name": "data-description",
      "content": "The data file contains three variables: trainx, trainy, testx.\n\n- trainx is a 463715×90 matrix of training data, where each row is a 90-dimensional feature representation of a song.\n- trainy is a 463715×1 vector of labels associated with each training data. Each row is the release year of the corresponding song in trainx.\n- testx is a 51630×90 matrix of test data.\n\nSpecial note for those who are not using Matlab: Python users can use scipy to read in the mat file, R users can use R.matlab package to read in the mat file.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "Welcome to COMS 4771 Fall 2021 large-scale regression task. Your goal for this assignment is to develop a good quality regressor for the given dataset.\n\nThe given dataset comes from musical songs domain. It consists of about 1/2 Million songs from 20th and 21st century. The task is to predict the year the song was released given the musical content of the song.\n\nTo ease the preprocessing burden, each song (observation) has already been vectorized into 90 high quality timbre based features.\n\nYou can use any resource/code/package/toolbox that is publicly available to develop your regressor. Your primary goal for this assignment is to develop a good quality song-year predictor. (This of course will be only possible if you employ good Machine Learning practices).",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "##Evaluation Criteria\nYou must get a Mean Absolute Error (MAE) score of at most 6.500 on the private holdout test-dataset to get full credit. If your MAE score is over the simple baseline (see leaderboard), then you will get zero credit. If your MAE score is between the simple baseline and the full credit threshold, then you will be awarded points linearly proportional to your deviation from the full credit threshold. For example, if your MAE is halfway between the simple baseline and full credit threshold, then you will get half credit. Extra points will be awarded to the highest performing students, using the following scoring mechanism:\n\n$$\\text{Extra Credit Points}= 5(1−\\frac{\\text{Your Rank}−1}\n{\\text{Number of Participants with MAE < 6 }}\n)⋅1[\\text{Your MAE}&lt; 6]$$",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "- You must design your regressor individually. Teams of more than one student are not permitted.\n- You must use your UNI as your \"team\" name in order to get points.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Submission",
      "content": "## Submission Format\nThere are 24000 test examples. The submission file will be a CSV file with two columns: `dataid` and `prediction`. `dataid` is a unique identifier (between 1 and 2400) for each ride, and `prediction` is the estimated duration that ride.\n\nFor example, if you predict trips 1 through 5 as lasting 600 seconds and the rest of the songs 6 through 51630 as lasting 30000 seconds the submission file must look as follows\n```\ndataid,prediction\n1,600\n2,600\n3,600\n4,600\n5,600\n6,30000\n7,30000\n8,30000\n9,30000\n10,30000\n11,30000\n...\n2399,30000\n2400,30000\n```",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}